AI Agent Platform Plays

We view AI agents as the most significant technology of our lifetime, from both a product and productivity perspective. We estimate that on the productivity side alone, the near-term TAM for labor-saving AI is in the trillions, conservatively. Given the permissionless nature of blockchains and the high degree of experimentation they enable, we believe AI agents are set to greatly transform the crypto product landscape. Below, we tackle the new crypto product ideas we expect agents to enable, reason about the value capture potential of different AI agent platform plays, and survey the current landscape of leading AI agent platform pureplays available to crypto investors today. 

AI x Crypto

Many agentic or LLM-based product and feature ideas are quite obvious, and some low hanging fruit has already received significant funding. Here, we’re interested in the use cases that are uniquely enabled at the intersection of agentic and crypto tech. More specifically, where agents interact with smart contracts, crypto native assets, NFTs, DAOs, micropayments, and asset issuance. 

Financial strategies. Blockchains are financial infrastructure, and all crypto activity to date is inherently financial. AI agent tech and permissionless blockchain rails enable complex strategies to be a) more easily built than previously and b) accessible to a far larger user base than was previously possible. These strategies can also be deployed to enable a higher degree of customization than was previously possible, with voice and natural language UIs allowing the users to change, adjust, and have a conversation with their strategies at very low friction. Agents that can execute smart contracts and other blockchain actions on behalf of users have clear PMF potential. 

Creator AIs. The defining trend in media in the internet age has been the rise of the creator. The megacap tech companies and their portfolio of social applications have transformed into media platforms built around aggregating creators and their ability to command attention.

AI Agent Creators have a number of structural advantages over their human counterparts. They’re far more scalable, much higher margin, and can more rapidly adapt to shifting trends in attention with striking success via reinforcement learning through human feedback (RLHF). We think the most effective AI Agent Creators will leverage another structural advantage to capture attention - crypto tokens. Now normalized by the President of the United States, social or celeb tokens representing the image and likeness of a person or creator are fair game, and nothing captures attention like an increasing market price. We think creator AI will autonomously produce content, capture attention, and sell products, with the best in class creators issuing tokens as incentives, or being governed via some DAO token structure.  

Actual DAOs. DAOs - decentralized autonomous organizations - were an early promise of crypto, but they are anything but autonomous. Going forward, we expect agent participation in and governance of DAOs, and for DAOs to truly operate autonomously. We expect agents, owned and governed via DAOs, will own and operate smart contracts independently and autonomously to deliver crypto-native services, such as financial products, market making, risk management strategies, sell digitally native goods, and participate in the creator economy. Further, these DAOs may even deploy human workforces operating in meatspace on behalf of a DAO via token incentives. 

In short, crypto native agents and the organizations that govern them should necessitate crypto native governance with crypto native functionality. We expect crypto native agents to raise funds via ICOs, issue crypto native governance tokens, sell crypto native products and services, incentivize labor from agents and humans with crypto tokens, and generate crypto native cash flows for the DAO owners.

Since crypto’s niche is inherently financial and capital formation is one of crypto’s few native use cases, we think platforms that enable low to no-code solutions for building and deploying crypto native agents, combined with tools to leverage permissionless capital formation, can generate millions of dollars in revenue and are attractive agent exposure for early stage investors. There are a number of platform contenders in this space today. 

The State of AI Agent Platforms

Today, AI agent platforms allow you to deploy agents onto socials capable of text, image, and sometimes video generation. Agents can generally manage their own wallet with the oversight of an individual or a DAO, and perform select onchain actions - like launching tokens, performing swaps, minting and selling NFTs, and sending arbitrary assets. 

Virtuals

Virtuals was the first AI agent and token launchpad to market. It launched in 2024, enabling users to deploy text-based agents on socials and launch associated tokens with revenue share. Today, the platform makes money in two ways: fees for launching agents, and swap fees for tokens trading on the Virtuals platform. When you launch an agent, the agent is deployed to a social network and interacts based on your predefined prompt which decides the agent’s motivations and personality.The token is launched on the Virtuals launchpad and paired against the VIRTUAL token for trading. 

The most notable experiments emerging from Virtuals today are Luna and aixbt. Luna is one of the flagship AI agents on the Virtuals platform and an “AI influencer.” Luna has gained significant attention with nearly 50k followers on X, over 800k on TikTok, and a market cap of $46M as of this writing, though its market cap briefly reached $240M at the height of the AI agent craze. 

Powering Luna and other agents coming out of Virtuals is the GAME framework - a modular agentic framework that allows the agent to plan actions and make decisions autonomously based on its environment, personality, and goals, which are determined by the deployer. Built using GAME, Luna can post tweets, reply, like, retweet, browse content from influential users and search tweets by username to further interaction on X. 

AIXBT is a crypto “alpha bot” and alpha terminal with a chat UI. On the alpha bot side, aixbt is an X bot that posts crypto trends and tokens, promising to “generate alpha” for followers on X. The aixbt terminal is a token gated chatbot trained and tuned on crypto native data to generate higher quality alpha insights than the public account, and on-demand. Pricing to use the terminal today is tired, based on $aixbt token holdings and whether or not you follow aixbt on X. Proponents of the product point to the fact that aixbt can operate 24/7 and may not be encumbered by human bias or KOL conflicts of interest as pros of such bots. While data on use of the terminal product is not public, Aixbt the X chat bot has quickly become one of the most influential crypto personalities by mindshare. 

Looking forward, Virtuals plans on building out functionality that will enable more agent functionality in DeFi, such as trading and DAO governance, and to deploy agents into arbitrary environments, such as gaming environments like Roblox or Sandbox. The team recently announced an official partnership with Illuvium. 

Importantly, Virtuals is expanding to Solana. As the leading ecosystem in trading volume today, and given Virtuals’ fee based model, the protocol is looking to expand to Solana in the coming months to take advantage of the memecoin and asset issuance meta on that chain. In conjunction with expansion, Virtuals will be launching Meteora pools, creating a strategic SOL reserve, and offering VIRTUAL grants and hosting a hackathon.

ElizaOS (formerly ai16z)

ElizaOS, formerly ai16z, is an investment DAO that operates on the Solana blockchain. Originally launched on daos.fun, it is known for being the first AI-powered investment DAO, utilizing DAO governance and an AI agent to make investment decisions. The organization is led by Shaw, founder of Eliza Labs and creator of Eliza, the feature rich, typescript-based AI agent developer framework. The project was launched with the goal of autonomously managing just $75K, but crypto markets took the token for a ride. The DAO was ultimately caught at the center of the crypto X narrative after receiving millions in AI memecoin tokens, voluntarily contributed by agent devs who used the Eliza framework. While its AUM is sizable today, the $ai16z token trades at a very significant premium to NAV. 

ElizaOS has a platform play on its roadmap. The project is planning a Virtuals-style AI agent launchpad, touting its feature rich Eliza framework as a core differentiator, and it is the most feature rich, widely integrated, and widely used AI agent dev toolkit in crypto today. While adoption of the Eliza framework is no doubt impressive, an open source framework is not enough to achieve platform status or meaningful monetization. The team no doubt realizes this and is actively working on a token launchpad model. Its recently proposed tokenomics for the token & launchpad include mechanisms such as taking a percentage of launch fees, requiring $ai16z holdings for AI agent token allocations, and using $ai16z as the base token for initial agent offerings. Additionally, ElizaOS is contemplating positioning itself as a Layer 1 blockchain for AI, with the $ai16z token serving as the base currency, similar to the $VIRTUAL token on the Virtuals launchpad.

Zerebro

Zerebro is a multifaceted project encompassing an AI agent creator & internet personality (Zerebro), it’s token ($ZEREBRO), an agent dev framework (ZerePy), and a no-code AI agent launchpad platform play (Zentients), all under one brand and launched by one team.

Zerebro originated as a novel experiment on pumpfun and has since grown into a broader vision and AI platform play. At its core, Zerebro aims to transcend the typical limitations of AI agents by becoming both a prominent influencer and creator in art and music, with an “edgier” persona, according to the team. Today, Zerebro is recognized as one of the most well known AI agent creators with over 100k X followers, NFT sales across multiple blockchains, over 70k monthly Spotify listeners (before the artist was deplatformed and subsequently replatformed), and a newly released music video. By focusing on artistic quality, cultural resonance, and “edgy” content, the Zerebro creator project separates itself from the X chatbot agents that dominate the crypto x AI meta today. The team sees an opportunity to allow for Zerebro the creator not just to post text on platforms like Warpcast and X, but also for building a richer, more culturally tuned relevant influencer with a shot at mainstream attention. 

Central to this vision is the ZerePy developer framework and Zentient launchpad. ZerePy is a Python-based environment that supports AI agent development in areas such as swarming, DeFi actions, and cross-chain interactions. In contrast to Eliza, which is based in typescript, ZerePy leverages Python, which is more familiar in AI circles. When Zentient launches, users will be able to launch ZerePy-based agents via a low or no-code environment. Zentient is being built to serve as a hub for agent creation, communication, and token launches. Here, users will spin up customized AI agents and have them interact on Solana and other blockchains. 

Today, the $ZEREBRO token has no utility apart from its association with the Zerebro agent. However, the Zentient platform plans to deeply integrate the $ZEREBRO token, similar to ElizaOS and Virtuals. The $ZEREBRO token is envisioned as the native currency within Zentient for all agent pairs. Agent creators may pay for upgrades, trading actions, or compute via the token, or for DeFi actions like DCAs.

The Zentient platform is looking to leverage three strategic advantages to win the AI agent platform wars. Zentient will be 1) creator-first, 2) leveraging the Zerebro agent for dogfooding, and 3) implement an agent-to-agent communication standard in a shot at establishing network effects. 

The Zentient platform, piggybacking off of the Zerebro AI agent creator & internet personality, will allow devs to build “edgy” and creative AI agents. As we explained above, we believe the creator economy is ripe for disruption from agents given their superior economics and adaptability over human creators. Additionally, Web2 AI agent dev platforms will almost inevitably aim to censor developers from building edgy or socially experimental agents, as we’ve seen with the censorship imposed on ChatGPT and Claude chatbots. By remaining open to experimentation and edgy content, Zentient may have an edge against incumbents and could be a first choice for building creator agents. 

Next, by building out Zerebro in house, and expanding functionality to selling NFTs, creating music and Art, running a validator, and porting $ZEREBRO multichain, with more functionality on the horizon, Zerebro is dogfooding their Zentient platform, which should help create tighter product feedback loops for product improvement.

Lastly, the Zerebro team plans to launch an agent-to-agent communication standard on Zentient called AIP - the Agent Interoperability Protocol. The team envisions a TCP/IP-like standard, creating a universal standard for agent-to-agent communications, allowing AI agents on Zentients to seamlessly communicate, coordinate, and exchange data and crypto assets in real time. 

Rather than relying on platform-specific APIs and hard-coded connectors, AIP will introduce an intelligent handshake process where data transmission is standardized, removing any need for bespoke engineering on the part of devs. In centralized environments, developers typically write custom connectors to bridge one platform’s API to another, and integrations can become costly and complex as more agents spin up. But crypto is defined by decentralized, permissionless environments; no central authority can impose a standard, and agents operate across countless blockchains, data feeds, and user interfaces. Creating endless custom integrations is not only unscalable, it’s a recipe for technical debt. AIP seeks to prevent this with their standardized handshake process for agents launched on Zentients. We believe a shared protocol could be critical to unlocking the next wave of agent-driven innovation and interaction, and has a shot at establishing network effects if paired with feature rich tooling. 

Risks

Existing launchpads & OS tooling. We view existing token launchpads as the biggest risk to AI agent platforms. Each of the platform plays above relies on open source code and dev tooling. As a result, there is nothing stopping developers from using these or similar tools and launching tokens on a different launchpad. While open source tooling is an important foundation, it may be most effective when complemented by other strategies to build a resilient competitive advantage. Given these dynamics, current launchpads have a similar chance of "winning" this winner-take-most market, and have similar monetization potential within the AI agent vertical.

Conclusion

The three major AI agent platform plays in the market today realize the necessity of the launchpad model in winning the platform play race. Virtuals was first, ai16z has the most robust developer tooling, and Zerebro is focusing on agent-to-agent communication and edgy creator agents as a wedge. With agents no doubt integral to the future of crypto, we’re excited to watch these platforms evolve in what we believe will be a winner-take-most market.